Transformation of n-heptane using an in-liquid submerged microwave plasma jet of argon

The reforming of fuel has been of special interest ever since the Fischer–Tropsch process, i.e., the liquefaction of syngas from coal, was first introduced in the early 20th century. Recently, the study of the reforming of fuel has diversified, partly in response to climate change issues, such as using CO₂ as feed or producing carbon-neutral (or carbon-free) fuels, and partly to cope with intermittent renewable electricity storing into chemical energy.

Since the early 2000s, nonthermal plasmas have been considered as a promising method for the on-board production of hydrogen-rich syngas, particularly because plasma devices are compact and respond quickly.¹ Dry reforming, partial oxidation, and steam reforming of methane have been popular chemical processes in the context of plasma-based fuel reforming.^2–5 Recently, much attention has been given to modeling the fundamental underlying plasma chemistry,^5–8 and recent experiments^9–11 have helped to identify the respective roles of energetic electrons and gas temperature in plasma chemistry; electrons initiate the chemical process by creating radicals, and the temperature determines the product distribution.

Recent advances in the field of in-liquid discharges have inspired the exploration of the reforming of liquid hydrocarbons.^12–16 These approaches practically do not require the evaporation of liquid fuels; thus, these can accommodate various kinds of feedstock and save the energy for the evaporation as compared to gas phase plasma methods. Studies have shown that the product composition of in-liquid plasma processes to transform liquid hydrocarbons can be tailored by adding chemical additives.^18–20 Particularly, methane and carbon dioxide were found to selectively increase the production of lighter hydrocarbons and oxygenates, respectively, in the reforming of n-dodecane¹⁹ and n-heptane.²⁰ Thus, there exist at least two ways to control the product composition: by controlling the reaction temperature and the choice of chemical additives.

n-Heptane has been considered as a model liquid fuel; therefore, its reforming has been extensively studied. Pyrolysis-induced n-heptane transformation is one of the popular topics in this subject area. Bajus et al.²¹ have proposed one of the well-known mechanisms, which is based on the Rice–Kossiakoff theory. The unimolecular scission of primary and secondary C–C bonds (C₇H₁₆→ R + R′; R and R′ are alkyl radicals) and the cleavage of the C–H bonds to form heptyl radicals (C₇H₁₆→ H + C₇H₁₅) have been proposed as the two main initiation steps. Also, in the presence of small radicals, such as H, CH₃, C₂H₅, and C₃H₇, hydrogen abstraction from C₇H₁₆ can occur to produce C₇H₁₅ and other light saturated molecules (H₂, CH₄, C₂H₆, and C₃H₈). Then, further decomposition of C₇H₁₅ radical occurs via the following four pathways:

Pant et al.²² have studied the kinetics and product distribution during n-heptane pyrolysis. They have reported that the n-heptane decomposition could be represented by a first-order reaction and proposed a comprehensive reaction scheme to address the various ways of product formation. Wang et al.²³ have studied n-heptane reforming using gliding arc discharges. Gaseous n-heptane mixed with air has been reformed into hydrogen-rich syngas. Particularly, Lebedev et al.²⁴ have performed a numerical study for microwave discharges in a small bubble of n-heptane vapor. They have found that the role of electrons in the dissociation of n-heptane could be important only during a very short initial stage of time (∼1 ms), indicating that a main role of microwave plasma is thermal. On the other hand, Reddy and Cha²⁰ have investigated the effect of additives (CH₄, CO₂, and H₂O) on n-heptane reforming using a dielectric barrier discharge (DBD) reactor at room temperature, minimizing the role of plasma's thermal effect. They have shown that the additives could selectively increase lighter hydrocarbons or oxygenates in liquid phase.

Recently, we have developed a submerged microwave plasma jet (MWPJ)¹⁵ and a submerged arc plasma jet¹⁶ to overcome the following technical challenges in general in-liquid pulsed discharges: (i) a significant dependence of discharges on the liquid's properties, such as permittivity and electrical conductivity, and (ii) a limited, relatively low level of electrical power to the liquid. We have decoupled plasma generation from the liquid's properties by injecting a remotely produced plasma jet into the liquid and achieved higher electrical power by using an arc torch or a microwave plasma jet. These submerged plasma jet systems have been verified in water, which has a relatively high relative permittivity (∼80) and a wide range of conductivity by decomposing organic compounds.^15,16 The submerged plasma jet systems are required to be further applied to transform liquid hydrocarbons, which have very low permittivity (∼2) and negligible conductivity.

In the present study, we employed a submerged microwave plasma jet (MWPJ) to transform liquid fuel. Main motivation to adopt the MWPJ was to investigate the effect of hotter plasma on product distribution, which is expected to be very different from the result in the DBD reactor with additives.²⁰ Particularly, because the submerged configuration was designed to quench the plasma by ambient liquid, we expected to find another way of controlling a product composition. Argon was used as the feed gas to the MWPJ, and n-heptane was used as the liquid fuel. To focus on the thermal effect of the MWPJ by maintaining the chemistry as simple as possible, we considered no additives other than Ar. We conducted a product analysis to obtain a conversion of n-heptane and product selectivities. In particular, we focused on the product selectivities to highlight contrasts between the submerged MWPJ and other results, such as pyrolysis and chemical equilibrium composition. To estimate gas temperature and electron density in the MWPJ, a spectroscopic analysis of light emission was also conducted. As a result, we presented a unique feature of the plasma chemistry in the submerged MWPJ system.

A submerged MWPJ system to transform liquid n-heptane was employed, which consisted of a microwave power system and a quartz reservoir with a MWPJ inlet (Fig. 1). A solid-state microwave generator (Sairem, GMS200W) delivered continuous microwaves at 2.45 GHz to a surfatron device (Sairem, Surfatron60) via a coaxial cable up to 200 W of adjustable microwave power, P_dis, which was estimated by substracting the reflected power from the incident power. A quartz tube with an outer and an inner diameter of 6.0 and 4.0 mm, respectively, was inserted at the center of the surfatron to guide the plasma gas. The surfatron was water-cooled, and the outer part of the quartz tube was air-cooled. Argon (99.999% purity) was selected as the plasma gas, which was supplied through the quartz tube. The flow rate of argon, Q_Ar, was varied in a range of 1–5 l/min using a mass flow controller. One end of the quartz tube was welded to the quartz reservoir (4.6 × 4.6 cm² square cylinder), which was filled with n-heptane (V_hep = 200 ml, 9.45-cm height in the reservoir), such that the Ar MWPJ was ejected into the liquid n-heptane directly. A gas outlet was placed at the upper part of the reservoir. We conducted the experiment in atmospheric pressure.

We investigated the physical characteristics of the submerged MWPJ using optical emission spectroscopy (OES). Particularly, in order to estimate the gas temperature, the rotational temperature of C₂ was calculated by fitting the experimental spectrum of the C₂ Swan band (Δν = 0) using the method developed by Cardoso et al.¹⁷ A spectrometer (Princeton Instruments, SP2750), which was equipped with a grating (900 grooves/mm, blazed at visible) and an intensified charge-coupled device (ICCD) camera (Princeton Instruments, PI-MAX3), was used to acquire the spectral data of the emissions. The exposure time for each acquired spectrum was 1 ms, and 50 spectra were averaged for each measured condition. A fiber optics was installed at the outside of the reservoir to remotely collect the light emission from the inlet of the MWPJ to the reservoir, and the measuring volume was 10 mm in diameter. Thus, we obtained temporally and spatially averaged spectra. A high-speed camera (LaVision, Image Pro HS) was used to investigate the dynamic behavior of the plasma–bubble interactions. A light-emitting diode lamp (Luminus, SST-90) illuminated the reservoir from the opposite side to the high-speed camera to clearly illuminate the interface between gas and liquid. The exposure time and the framing rate of the high-speed imaging were set to 0.47 ms and 2 kHz, respectively.

We evaluated the chemical effects of the submerged MWJP on the transformation of n-heptane based on the resulting chemical products. The gas products were analyzed using a gas chromatography, GC (Agilent, GC 7890A), and the liquid products were identified by gas chromatography-mass spectrometry (Agilent, GC-MS/FID 7890A-5975C). The total plasma processing time, $tproc$⁠, with MWPJ was maintained at 5 min for all tested conditions. Note that, at P_dis = 200 W and $tproc=5min$⁠, the increase in the temperature of liquid n-heptane was ∼40 K, which may have negligible influence on the chemical reaction. The chemical analysis was repeated five times for each experimental condition, and the results showed less than ±3% deviation from the averaged concentrations for each species. Thus, we plotted the data based on the averaged concentrations.

We investigated the transformation characteristics of the submerged MWPJ to liquid n-heptane with particular emphasis on the conversion of n-heptane and the selectivities of the produced components. To assess the conversion and the selectivities from the analytical data obtained from the GC and the GC-MS analyses, we developed the following equations.

For a species x in the gas products, its outlet molar flow rate, $n˙x,gas$⁠, can be expressed as $n˙x,gas=Cx,gas⋅(n˙Ar/CAr)$⁠, where $n˙Ar$ denotes the inlet molar flow rate of Ar. $Cx,gas$ and $CAr$ denote the outlet concentrations of x and Ar measured by the GC, respectively. The total produced moles of x, $nx,gas$⁠, can be calculated as $nx,gas=n˙x,gas⋅tproc$⁠. Thus, the total amount of carbon atoms (⁠$NC,gas$⁠, mole-based) in the gas products can be obtained from

where i is the number of carbon atoms in the species x.

For the remaining liquid in the reservoir, which included unconverted n-heptane, the total amount of carbon atoms (⁠$NC,res$⁠, mole-based) is calculated using a carbon balance as $NC,res=7⋅nC7H16−NC,gas$⁠, where $nC7H16$ is the initial molar amount of n-heptane in the reservoir. Thus, the produced moles of y species (⁠$ny,liq$⁠) in the liquid products can be estimated by $ny,liq=(NC,res⋅Cy,liq)/j$⁠, where j is the number of carbon atoms in the produced liquid species y, and $Cy,liq$ is the carbon-based concentration of the product y obtained from the GC-MS/FID analysis. Thus, the total amount of carbon atoms (⁠$NC,liq$⁠, mole-based) only as the form of transformed liquid, except for unconverted n-heptane, can be calculated as

From Eqs (1) and (2), the molar amount of converted n-heptane (⁠$nC7H16,conv$⁠) and the rate of n-heptane conversion (⁠$Xhep$⁠) can be obtained as

In the gaseous product, we identified H₂, CH₄, C₂H₂, C₂H₄, C₂H₆, C₃H₆, C₄H₈, C₃H₈, C₄H₁₀, and C₅H₁₀, whereas in the liquid product, we found C₄H₈, C₅H₁₀, and C₆H₁₂. Note that no solid particles were found. The selectivity of the produced H₂ (⁠$SH2$⁠) is given by

For a gaseous and liquid hydrocarbon product x and y having i and j carbon atoms, respectively, their selectivities based on carbon atom, $Sx$ and $Sy$⁠, are

Based on these calculations, we found that the energy density (or specific input energy) is not a useful parameter to explain n-heptane conversion in the submerged MWPJ system, unlike most of the hydrocarbon reactions in gaseous discharges that have been described in previous studies.^9–11 Energy density is defined as a density of the delivered electrical energy into a control volume and, in this study, can be defined for two different control volumes: one is the liquid n-heptane and the other is the flowing argon. The energy density for the liquid n-heptane, ED_hep = P_dis⋅t_proc/V_hep, may play an important role in characterizing X_hep in a similar way to the decomposition of organic compounds that are dissolved in water using a submerged arc jet system.¹⁶ 

The conversion we employed here seemed to be reasonably proportional to ED_hep when P_dis is varied at Q_Ar = 2 l/min. Meanwhile, X_hep was significantly affected by Q_Ar, which does not affect ED_hep [Fig. 2(a)]. We found one order of difference as X_hep = 20.2 and 2.77 μmol/s at Q_Ar = 2 and 5 l/min, respectively, for fixed P_dis = 200 W. This indicates that the effectiveness of P_dis to X_hep is influenced by the quality of the MWPJ and the interaction between the MWPJ and the liquid.

Although the energy density in the argon flow, ED_Ar = P_dis/Q_Ar, was considered instead of ED_hep, we found no generalized trend for the conversion of n-heptane in terms of ED_Ar [Fig. 2(b)]. A strong correlation of X_hep with increased ED_Ar for the varying P_dis conditions at Q_Ar = 2 l/min was still found. However, for the varying Q_Ar cases at P_dis = 200 W, X_hep showed a maximum at ED_Ar = 6 kJ/l (Q_Ar = 2 l/min), with unchanged X_hep in the range of 4 ≤ ED_Ar ≤ 12 kJ/l (1 ≤ Q_Ar ≤ 3 l/min).

Interestingly, we found the selectivities of the products to be well correlated with the ED_Ar, as highlighted by the shaded area in Fig. 3. We identified H₂, alkane (CH₄, C₂H₆, C₃H₈, and C₄H₁₀), alkene (C₂H₄, C₃H₆, C₄H₈, and C₅H₁₀), and alkyne (C₂H₂) present in the gaseous product, whereas in the liquid product, we only found alkene (C₄H₈, C₅H₁₀, and C₆H₁₂). The submerged MWPJ was highly selective to produce C₂H₄ followed by C₂H₂, C₃H₆, C₄H₈, H₂, and CH₄. Note that although S_H2 (<7% for all tested conditions) was comparable with S_C4H8, we have not shown this information in Fig. 3 for the sake of visibility and clarity.

Thus, the well-correlated results in the shaded area raise an interesting point; the reaction temperature might strongly affect the product composition since ED_Ar, by definition, may be related to the gas temperature in the MWPJ system. More specifically, our results indirectly confirm that most of the reactions occurred in the gas phase with the evaporated n-heptane in the MWPJ. In our previous studies on the basic reforming processes of methane—partial oxidation, dry reforming, and steam reforming—using a temperature-controlled dielectric barrier discharge (DBD) reactor,^9–11 we found that the gas temperature mainly controlled the product selectivity, and the reactant conversion was initiated by electron-induced chemistry, depending on electron temperature and density.

However, our selectivity results at two conditions, P_dis = 200 W:Q_Ar = 5 l/min and P_dis = 50 W: Q_Ar = 2 l/min, showed significant deviation from each other. The selectivities of C₂H₄, C₂H₂, and CH₄ abruptly dropped as Q_Ar increased to 5 l/min from 4 l/min at P_dis = 200 W (from ED_Ar = 3 kJ/l to 2.4 kJ/l) resulting in increased $SC3H6$ and $SC4H8$⁠. These results indicate that the effect of the argon flow on the selectivities and conversion is complicated. Thus, we provide in-depth discussion in Secs. III B and III C by elaborating on the physical characteristics of the MWPJ and detailing the role of ED_Ar on the selectivities, respectively.

We characterized the submerged MWPJ using optical emission spectroscopy to investigate its composition, temperature, and electron density. Based on the spectra obtained at P_dis = 200 W and Q_Ar = 2 l/min (Fig. 4), we identified the spectral lines of Ar (in the range of 700–850 nm), C₂ (Swan: head bands at ∼474, 517, and 564 nm and C₂-Deslandre-D′Azambuja bands bands: head bands at ∼361, 385, and 410 nm), CH (A-X: head band a ∼431 nm, B-X: head band at ∼387 nm, and C-X: head band at ∼315 nm), Hα (at 656.3 nm), and Hβ (at 486.1 nm). NH (A-X: head band at ∼336 nm) was also detected, which we attributed to impurities in Ar or to dissolved air in n-heptane. The observed emission spectra indicated that the initial n-heptane was successfully decomposed by the plasma to hydrogen radicals and CH radicals. Note that since n-heptane is colorless, there is no significant interference in the visible range.

As a first approximation, the estimated rotational temperature of C₂ might be stretched to represent the gas temperature (T_g) for a given condition.^25,26 Figure 5 shows a typical calculated synthetic spectrum, which fits for the experimentally measured C₂ Swan band (Δν = 0), to estimate the rotational temperature of C₂. We found a good agreement between the simulated and the experimental spectra for all tested conditions; as a result, the estimated gas temperatures were ranged in 1915–2732 K and 1683–2467 K for fixed P_dis = 200 W and fixed Q_Ar = 2 l/min cases, respectively. Note that the synthetic spectra showed a margin of uncertainty at ±50 K, as we found the best fit with the experimental one.

The energy balance might provide a simple explanation for the increase in gas temperature. Although P_dis is dissipated into chemical formation, photon emission and the sensible enthalpy of both gas and liquid, a proportionality between the input power and a resulted temperature change is still valid: P_dis∼ ρQ_Arc_pΔT_g, where ρ is the gas density, c_p is the specific heat, and ΔT_g is the temperature difference. Therefore, ΔT_g ∼ P_dis/Q_Ar can be obtained explaining inverse proportionality of the temperature to Q_Ar and its linear proportionality to P_dis. Recalling that the parameter P_dis/Q_Ar = ED_Ar, we plotted the estimated gas temperatures in terms of ED_Ar in Fig. 6.

As a result, we found that the estimated gas temperature adequately represented ED_Ar. Thus, the discrepancies in the selectivities shown in Fig. 3 can be attributed to something other than the gas temperature in the MWPJ.

To determine the electron density, we considered Stark broadening of Hα and found that the influence of the electron density on the kinetics of n-heptane transformation was neutral. By fitting the experimentally obtained Hα line with a Voigt profile, we determined the Lorentzian broadening $(ΔλFWHM)$ induced by the Stark effect considering the broadenings induced by van der Waals, Doppler, and the spectrometer in the fit; thus, the electron density in the MWPJ could be estimated using Gigosos's table.²⁷ We found that the electron density was in the range of 2.5 × 10¹⁵– 4 × 10¹⁵cm⁻³ for all tested conditions. This range of electron densities seemed to be insignificantly effected by a fourfold increase in P_dis (50–200 W) and a fivefold variation of Q_Ar (1–5 l/min) and ED_Ar (2.4–12 kJ/l); thus, electron density's effect on the chemistry should be rather uniform throughout the tested cases. This indicates a less significant role of electron impact reactions in “hotter” plasmas, and a previous study²⁴ for n-heptane decomposition using an direct in-liquid microwave plasma with a metallic electrode has supported the importance of thermochemistry.

The large differences in the selectivities at Q_Ar = 5 l/min and P_dis = 200 W shown in Fig. 3 may be attributed to significantly altered plasma dynamics and interactions with the liquid. To explore the causes behind the exceptional selectivities at Q_Ar = 5 l/min and P_dis = 200 W (Fig. 3) as well as the n-heptane conversions for the cases with varying Q_Ar at fixed P_dis = 200 W (Fig. 2), we investigated the dynamic interaction between the MWPJ and the liquid. As shown in Fig. 7, for the cases with fixed P_dis = 200 W, increased Q_Ar resulted not only in a longer plasma channel, but also in a stronger interaction between the plasma and the liquid. At low Q_Ar, the plasma existed within a limited region of the confined bubble, and there seemed to be no direct interaction between the plasma and the liquid. However, at Q_Ar = 5 l/min, the void region surrounding the plasma channel was significantly reduced, showing a pinched plasma zone in the liquid. Therefore, we could expect a rapid quenching of the MWPJ due to a direct contact with liquid n-heptane. This quenching process might drastically modify the overall reacting conditions for the n-heptane transformation. Note that the varying P_dis conditions (bottom row in Fig. 7) did not show such a pinching phenomenon, demonstrating that the void gas region—a buffer between the plasma and the liquid—for all tested conditions mildly influenced plasma and bubble behavior due to increased P_dis.

We would expect that the gas temperature in the MWPJ has a primary effect on product selectivity regardless of Q_Ar and P_dis, since previous methane reforming studies^9–11 have shown that the gas temperature strongly influences selectivities. We replotted the product selectivities as a function of T_g in Fig. 8. Similar to the selectivity result with ED_Ar (Fig. 3), an overlap of the selectivities in a T_g range of 1950–2450 K was found between both experimental sets, i.e., with varying P_dis at Q_Ar = 2 l/min and with varying Q_Ar at P_dis = 200 W. These results were as we expected because the estimated gas temperature showed a strong positive correlation with ED_Ar, as shown in Fig. 6.

Results from the chemical equilibrium calculation suggest that a much lower temperature range is required to produce the experimentally obtained product composition. We calculated the chemical equilibrium based on the Gibbs free energy minimization method in MATLAB with pure n-heptane as the initial composition for a range of the equilibrium temperature, T_eq. With the obtained molar concentrations of the products at various temperatures, we calculated the selectivity of each product to compare with our experimental results. The chemical equilibrium dictated the product in a range of T_eq = 1950–2450 K to have only H₂ and C₂H₂, as shown in Fig. 9, where we observed more than 40% C₂H₄ and moderate levels of C₂H₂ and C₃H₈ in the submerged MWPJ system (Fig. 8). To match the experimentally observed product composition to the chemical equilibrium result, we found that 950 < T_eq < 1100 K would be the most probable range of temperature. Similar to the experimental result in Fig. 8, we found a peak of $SC2H4$ with high selectivity (∼30%) together with increasing $SC2H2$ and decreasing $SC3H6$ and $SC4H8$ as the temperature increased. Because there is a steep temperature gradient in the MWPJ due to significant heat loss to the liquid n-heptane, the overall reaction temperature could be reduced compared to the estimated T_g using light emission.

Furthermore, the unusual selectivities at P_dis = 200 W and Q_Ar = 5 l/min might also be due to severer quenching of the MWPJ compared to other Q_Ar, as shown in Fig. 7. Due to enhanced heat loss, caused by the direct contact with the liquid and turbulence mixing, the effective reaction temperature of this case should be further decreased compared to other cases. When we projected the temperature below 950 K, to 800 K for example, we obtained a significant drop in $SC2H4$ and $SC2H2$ and an increase in $SC3H6$ and $SC4H8$ in the equilibrium calculation (Fig. 9), similar to the result with the MWPJ (Fig. 8).

We must highlight a significant contrasting point in terms of CH₄ and C₂H₄. According to the chemical equilibrium in 950 < T_eq < 1100 K, methane is the most probable product; however, our MWPJ experiment showed less than 5% selectivity for CH₄, producing C₂H₄ with more than 40% selectivity. Previous chemical kinetic studies^28,29 have shown that, in a pyrolysis of n-heptane, the formation of CH₄ is much slower than that of C₂H₄. In this regard, when a reaction time is limited by a system, the formation of CH₄ is expected to be reduced. The present submerged MWPJ underwent spatial and temporal, rapid quenching of the reaction due to the cold liquid boundary. Thus, there might be not sufficient time for methane to form.

Note that studies for a thermal dissociation (pyrolysis) of n-heptane, both experimentally³¹ and theoretically,²⁸ have found that C₂H₄, H₂, CH₄, C₃H₆, and C₂H₆, and often H₂ and C₂H₄ were the most dominant products, with barely any report on the presence of C₂H₂. Meanwhile, C₂H₂ is the second populous species in the present study. This might be attributed to that the effective reaction temperature of the present experiment happened to correspond to the favorable range of temperature for the formation of C₂H₂ production as shown in Fig. 9. However, as one-dimensional modeling study³⁰ for microwave plasma bubble in n-heptane has shown a highly dynamic nature of product distributions in time and space, a detailed modeling study should be required to understand the underlying plasma-chemical effect.

Therefore, we concluded that the observed consistent selectivities, irrespective of various experimental conditions, indicate the dominance of related thermochemistry for the tested ranges of power and flow rate. In addition, due to the nature of the submerged MWPJ, high-temperature jet, and rapid quenching of the chemical reaction, we could obtain the different product composition than that obtained by the pyrolysis and the chemical equilibrium.

We experimentally investigated a submerged MWPJ of argon in liquid n-heptane to study its transformation capability with liquid substance and to prove its unique character in controlling product selectivity due to the rapid quenching of chemical reactions caused by heat interaction with the liquid. We found that product selectivities are controlled by gas temperature, and thus, the energy density, as defined by the electrical power and the flow rate of argon, was the key parameter for this selectivity. Conversion of n-heptane was primarily proportional to the electrical power but was significantly coupled with the MWPJ and the liquid as the flow rate of argon increased. Comparing the experimental results with the chemical equilibrium compositions, the obtained product selectivity seemed to be governed by the related thermochemistry (pyrolysis). However, the primary products with high selectivities were ethylene and acetylene, which was in contrast to the methane or hydrogen/ethylene via the chemical equilibrium or the pyrolysis, respectively. We determined that this was most likely due to the nature of the submerged MWPJ—rapid quenching. Furthermore, the measured electron density also indirectly indicated that the role of electron-induced chemistry with the submerged MWPJ on the selectivity was insignificant compared to the related thermochemistry.

The research in this publication was supported by funding (No. BAS/1/1384-01-01) from King Abdullah University of Science and Technology (KAUST).

The data that support the findings of this study are available within the article and from the corresponding author upon reasonable request.

A.H. performed the experiment for the physical analysis. J.-L.L. performed the experiment for the chemical analysis. M.S.C. conceived the experiment and constructed the main idea of the study. All authors contributed to the preparation of the manuscript.